Correlation grouping

[nbt1032]

I have a large dataset of about 12 million records.  There are 2 factors that I would like to correlate.  Running this correlation on the entire set does work, but I would also like to see the correlation by a field grouping (i.e. State).

Does anyone have some ideas on how this can be done?

Thanks,

Marc.

[JordanB]

Hi @nbt1032

I have attached one potential way you can group by the State field. My example uses cities but you should be able to replicate. 

I used a batch macro and the grouping feature.

Best,

Jordan Barker

Solutions Consultant 

[nbt1032]

Thanks Jordan.  Little above my current skill level, but I am going to work through the file you sent to understand it.   I am an experienced developer, but new to Alteryx.

In my example, I would get a single correlation value.  Can you describe how this workflow would group those by some other factor?

I envisioned a table similar to below.  Would what you are suggesting create this.  I believe it would need to be an iterative process that would go through each state value and run the correlation logic.  I was hoping not to have to build a loop but use some sort of "group by" function.

State	R value
Michigan	.1
New York	.2
Florida	.3
Georgia	.4
Alabama	.5

[JordanB · Accepted solution]

Hi @nbt1032

The Batch process will process each record one by one that you feed into it, similar to a list.

Once it proceses that first record it will loop back and process the second record (or second city in my example)

You will then end up with a unioned table similar to what you have in you table above. 

The iterative loop is another type of macro process we can use in Alteryx similar to the developer concept where it loops untill a condition is met. In this instance it will just group by the value you pass in through the filter. Process those records and perform the correlation and then loop back and do the same process will the second record. 

Best,

Jordan Barker

Solutions Consultant

[nbt1032]

To get the correlations, I will need a set of records, so your are suggesting the way to do this, is to processes each set (in my case states) of records, one set at a time through a loop.  This is what I thought.

Thanks for the help.

Marc.

[Ashish]

The pearson correlation tool sometime ago had grouping option but it was depreciated and the new tool now does not have the grouping option, The upgrade I see is that we can select more than two variables but we already had it in "Association Analysis" tool. I think the tool which is now depreciated had more use than the new one.

Also, its annoying that Alteryx is now disabling the way we could open macros to customize them (now the new regression tools and Correlation tool are sealed. This is making them useless for us as we need results in certain different ways and it was much easier to just tweak a little bit  but now we feel stuck with what we have.